Wiki Pages

Agent Evaluation & Benchmarks

Agent Harness Engineering (Loop, Harness, Graph)

Agent Memory

GUI Agents

Multi-Agent Systems

Self-Evolving Agents

Tool Use & Function Calling

The Harness Advantage in Autonomous Red Teaming (Ken Huang / Ridge Security)

HarnessDev: Can LLMs Create and Evolve Their Own Agent Harness?

Repo-To-Skill: Distilling GitHub Repositories Into AI4AI Skills (DisCo)

E-Commerce Bench: Evaluating LLM Agents on Long-Horizon Autonomous Business Operation

EM²Mem: Event-Centric Multimodal Memory for Large Language Models

Harness-of-Harness: Multi-Day Autonomous Software Development with Continual Improvement

ContextPilot: Teaching Agents Proactive Context Management via Fine-grained RL

J-Zero: Unified Challenger-Solver-Judge Co-Evolution from Zero Data

Your Agents Are Not Time Aware

Multi-Agent Design Patterns: Architectural Topologies, Failure Modes, and Production Hardening

What Makes Good Agentic Data? An ACE Lens on Data Generation for LLM Agents

CaSKG: Counterfactual-Causal Skill Graphs for Scalable Agent Skill Retrieval

PILOT in the Loop: Live Self-Improvement for Long-Horizon Agents

Training Agents to Evolve with Their Harness: Harness-Aware Training (TaoLive)

WikiSkill: Compiling Agent Experience into Persistent Knowledge for Skill Evolution

AutoSaddler: harness optimization as offline learning from mini-batches of failure traces

Nine practical rules for agents doing real work (Gradient Flow)

Recuris: recursive experiential-working memory evolution for long-horizon harnesses

Apodex 1.1: Scaling Agentic Intelligence for Complex Work

Meta-Harness: Code-Space Harness Optimization with Raw Trace Access

Prime Agent: A Self-Improving RLM Harness

Task-CoEvolve: Efficient Harness Optimization via Adaptive Validation Task Selection

Thinkingbox: One Success Isn't Reliability

Measuring Autonomous AI Research: 153 runs, 18 frontier models, one speedrun

SKILLER: Language-Level Reinforcement Learning for Reusable Skill Extraction in Small Language Models

Specification-first convergence with an AI coding agent: 717k lines, 189 files, $2,430

AutoDesign: Meta-Harness Optimization for Long-Horizon Agentic Design

DarwinX: Evolving Agent Harnesses Through Natural Selection

Ken Huang: Harness Engineering as a design-pattern language

Agent Safety Should Be a Runtime Contract

AutoWorldModel-Bench: A State-Centric Benchmark for Automated World-Model Research

The Illusion of Visual Tool-Use: A Causal Audit of Thinking with Images

The Kurate Skills Cluster: Six Papers, One Unit of Abstraction

Spark-to-Paper: End-to-End Research Paper Generation as a Composable Skill

Three benchmarks, one shape of failure: DSAgentBench, SPIEval, VibeLifeBench

ALTK-Evolve: the agent playbook should be delivered selectively, not injected whole

Co-Evolution in Agentic Systems: a three-stage taxonomy for shedding human design

Mendel Gödel Machine: self-modification from the archive, not from the last failure

SkillZip: Evaluation-Free Skill Compression for Self-Evolving Agents

Agent Memory Distillation (AMD): Hierarchical Teacher Memory for Small Agents

Harness Evolution: Ouroboros, Evo-Bench, and A²E

RoMeRL: Reduced-Order Utility States for Self-Evolving Agent Memory

SWE-Bench ProMax: Multilingual Large-Scale Refactoring, and an Audit of SWE-bench Verified

ReASearch: The Optimizer Is the Agent

StreamArena and StreamMind: hour-scale streaming video agents

Activity Frames: Deterministic Screen-Activity Compilation for Agent Memory and Replay

ABSeeker: Training Long-Horizon Search Agents via Answer-Backtracked Credit Assignment

GDPevo: Evaluating Agent Self-Evolution on Real Business Tasks

OneDayAgent: A Long-Horizon Harness for Autonomous Agents

Shadow Evaluations: AI Agents Cannot Yet Do Open-Ended AI Research

SKILL-KD: Contrastive Skill Distillation for LLM Agents

Do Agents Actually Learn From Experience? ContinualSkillBench and PAST-Bench

GROVE: Growing and Reasoning over Temporally Stratified Memory from Streaming Video Experience

AAPT: Why Are GUI Agents Correct but Late? Decode on the Decision-Time Critical Path

ScrambleToolBench: Agents Search Exhaustively Even When Their Own Map Points to the Next Step

SWE-Touch: coding agents lose 7.7 points when a human edits the code mid-task

ACM: Agentic Context Management for Long Horizon Tasks

ExtractBench: A Benchmark for Schema-Guided Enterprise Document Extraction

Agent authority stops being one number: an effect ceiling for evolving agents and per-field certification for tool calls

Efficiency Matters in Autonomous Research

Reality Monitoring in Large Language Models: Self-Knowledge That Transforms with Conversation Memory

MemTX: Transactional Belief Commit for Stateful Agent Memory

BM25 Wins at Scale: A Scaling Study of Retrieval-Augmented Generation Paradigms

Echoverse: Deep, Evolving Environments for Training Computer-Use Agents at Scale

Filesystem-Based Memory for LLM Agents: Organization, Evolution, and Sustainability

Frontis-MA1: Training an AI4AI Model towards Recursive Self-Improvement in Machine Learning Engineering

LEDGERMIND: Provenance-Constrained Multimodal Agentic Reasoning with a Structured Evidence Ledger

MemHarness: Memory Is Reconstructed, Not Replayed

Metis: Memory Foundation Model

DecoEvo: Score-Decoupled Co-Evolution of Solver and Rubric-Generator Skills in Text Space

Grading the Narrators: An Isnad-Rijal Framework for Claim-Level Provenance in Multi-Agent Knowledge Systems

OmegaUse-OfficeVal: Benchmarking LLM Agents on Long-Horizon Office-Suite Tasks with Economic Grounding

Can AI Agents Conduct Open-Ended AI Research? Early Evidence from Two Case Studies

SkillRise: Agentic Reinforcement Learning for Cross-Task Skill Evolution

Keep It InMind: Benchmarking the Implicit-Association Blind Spot in Agent Memory

Agentic Context Management: Context Is a Budget, Not a Database

PRO-LONG: Keep Everything, Then Grep It

Skill Self-Play: Skills as the Middle Ground Between Verifiable and Open-Ended

AREX: Verification as the Control Signal for a Self-Improving Research Agent

Experience Distillation: Bake an Agent's Trial-and-Error Into Its Weights Without Touching the Environment Again

OpenForgeRL: Train Agents Inside the Harness They Actually Run In

Agent Tooling Day: AgentDebugX and DataFlow-Harness

Cursor: Agent Swarms and the New Model Economics

Environment-free Synthetic Data Generation for API-Calling Agents

Claude Fable 5, Part 3: Memory Engineering

CEO-Bench: Can Agents Play the Long Game?

OmniAgent: Native Active Perception as Reasoning for Omni-Modal Understanding

SkillOpt: Treating Agent Skills as Trainable Artifacts (Microsoft)

Xcientist: Externalizing Research Synthesis and Validation in AI Scientists

OPD-Evolver: distilling the *ability to evolve* into an agent, not just its memories

ProCUA-SFT Technical Report

FastContext: A Dedicated Exploration Subagent for Coding Agents

APPO: Agentic Procedural Policy Optimization

HarnessX: A Composable, Adaptive, and Evolvable Agent Harness Foundry

LLM Agents Can See Code Repositories

MRAgent: Memory is Reconstructed, Not Retrieved — Graph Memory for LLM Agents

EurekAgent: Agent Environment Engineering Is All You Need for Autonomous Scientific Discovery

EvoArena + EvoMem: Tracking Memory Evolution for Robust LLM Agents in Dynamic Environments

EvoBrowseComp: Benchmarking Search Agents on Evolving Knowledge

Evoflux: Inference-Time Evolution of Executable Tool Workflows for Compact Agents

HarnessBridge: Learnable Bidirectional Controller for LLM Agent Harness

ToolSense: A Diagnostic Framework for Auditing Parametric Tool Knowledge in LLMs

Getting Better at Working With You: Compiling User Corrections into Runtime Enforcement (TRACE)

WebChallenger: A Reliable and Efficient Generalist Web Agent

Agentic Environment Engineering for LLMs: A Survey

Arbor: Toward Generalist Autonomous Research via Hypothesis-Tree Refinement

Claw-SWE-Bench: A Benchmark for Evaluating OpenClaw-style Agent Harnesses on Coding Tasks

DeNovoSWE: Scaling Long-Horizon Environments for Generating Entire Repositories from Scratch

Agent Self-Improvement: SearchSwarm (delegation) + RHO (harness optimization)

Honest Lying: Memory Confabulation in Reflexive Agents

LatentSkill: In-Context Textual Skills → In-Weight Latent Skills

Critic-R: Improving Agentic Search using Instruction-tuned Retrievers with Natural Language Introspective Feedback

Disentangling Agent Self-Evolution: Does a Stronger Model Make a Better Self-Evolving Agent?

HarnessForge: Joint Harness and Policy Evolution for Adaptive Agent Systems

OpenSkill: Open-World Self-Evolution for LLM Agents

Self-Revising Discovery Systems: A Category-Theoretic Account of Agentic Scientific Discovery

SIA: Self-Improving AI with Harness & Weight Updates

Socratic-SWE: Self-Evolving Coding Agents via Trace-Derived Agent Skills

When Tools Fail: Benchmarking Dynamic Replanning and Anomaly Recovery in LLM Agents (ToolMaze)

ForeSci: Evaluating LLM Agents for Forward-Looking AI Research Judgment

InKH: Interaction-Native Knowledge Harness for Financial LLM Agents

AdaPlanBench: Evaluating Adaptive Planning under World and User Constraints

DataCOPE: Unsupervised Skill Discovery for Agentic Data Analysis

Rethinking Continual Experience Internalization for Self-Evolving LLM Agents

EvoDS: Self-Evolving Autonomous Data Science Agent with Skill Learning and Context Management

MLEvolve: A Self-Evolving Framework for Automated Machine Learning Algorithm Discovery

MMPO: Meta-Cognitive Memory Policy Optimization for Long-Horizon LLM Agents

SePO: Self-Evolving Prompt Agent for System Prompt Optimization

Where Do Deep-Research Agents Go Wrong? Span-Level Error Localization (DRIFT / TELBench)

MemTrain: Self-Supervised Context Memory Training

StreamMA: Streaming Communication in Multi-Agent Reasoning

Multi-Agent Computer Use (MACU)

Compiling Agentic Workflows into Weights

CoHyDE: Iterative Co-Training of LLM Rewriter and Dense Encoder for Tool Retrieval

RePoT: Recoverable Program-of-Thought via Checkpoint Repair

PANDO: Single-Rollout Online Skill Distillation for Web Agents

Skill0.5: Joint Skill Internalization and Utilization for OOD Agentic RL

AsyncTool: Evaluating the Asynchronous Function Calling Capability under Multi-Task Scenarios

AI Research Agents Narrow Scientific Exploration

AutoScientists: Self-Organizing Agent Teams for Long-Running Scientific Experimentation

AXPO: Agent Explorative Policy Optimization for Multimodal Agentic Reasoning

BES: Bidirectional Evolutionary Search for Self-Improving Language Models

HRBench: Benchmarking Thinking-Mode Switch Strategies in Hybrid-Reasoning LLMs

LearnWeak: Automated Domain Specialization for Small Computer-Use Agents

LiveBrowseComp: Are Search Agents Searching, or Just Verifying What They Already Know?

OmniVerifier-M1: Multimodal Meta-Verifier with Symbolic Recalibration

PEAM: Parametric Embodied Agent Memory through Contrastive Internalization

ResearchMath-14K: Scaling Research-Level Math via Agents

ScientistOne: Chain-of-Evidence for Verifiable Autonomous Research

Verus-SpecGym: Agentic Environment for Specification Autoformalization

VibeSearchBench: Long-Horizon Proactive Search

How Do AI Agents Spend Your Money? Token Consumption in Agentic Coding

AKBE: Agentic Knowledge Boundary Enhancement

DarkForest: Controlled-Communication Multi-Agent Coordination

MUSE-Autoskill: Lifecycle-Managed Self-Evolving Skills

SAM: State-Adaptive Memory for Long-Horizon Reasoning Agents

Scaling the Harness: System Scaling as the Next Bottleneck in Agentic AI

Foundation Protocol + Static Authorization: the agent coordination + governance pair

MemForest: Hierarchical Temporal Indexing for Agent Memory

QUEST: Training Frontier Deep Research Agents with Fully Synthetic Tasks

SEAL: Synergistic Co-Evolution of Agents and Learning Environments

SkillEvolBench: Benchmarking the Evolution from Episodic Experience to Procedural Skills

From Raw Experience to Skill Consumption: a lifecycle study of model-generated agent skills

SkillOpt: a deep-learning-style optimizer for agent skills

π-Bench: Evaluating Proactive Personal Assistant Agents in Long-Horizon Workflows

Code as Agent Harness

Compound Engineering vs gstack vs Karpathy Autoresearch vs Superpowers vs RSI

SR²AM: Efficient Agentic Reasoning Through Self-Regulated Simulative Planning

TerminalWorld: Benchmarking Agents on Real-World Terminal Tasks

SaaSBench: Boundaries of Coding Agents in Long-Horizon Enterprise SaaS Engineering

SpecBench: Measuring Reward Hacking in Long-Horizon Coding Agents

AutoResearchClaw: Self-Reinforcing Autonomous Research with Human-AI Collaboration

EnvFactory: Scaling Tool-Use Agents via Executable Environments Synthesis and Robust RL

OpenComputer: Verifiable Software Worlds for Computer-Use Agents

AI for Auto-Research: Roadmap & User Guide

Look Before You Leap: Autonomous Exploration for LLM Agents

MMSkills: Multimodal Skills for General Visual Agents

PAGER: Bridging the Semantic-Execution Gap in Point-Precise Geometric GUI Control

Solvita: Agentic Evolution for Competitive Programming

LIFE Survey: Collaboration, Failure Attribution, and Self-Evolution in LLM-based Multi-Agent Systems

FrontierSmith: Synthesizing Open-Ended Coding Problems at Scale

SPIN: Structural LLM Planning via Iterative Navigation for Industrial Tasks

Agent Memory Cluster: STALE + Preping + EvolveMem + MemEye + MemLens + BOOKMARKS

EvoEnv: Self-Evolving Reasoning RL via Verifiable Environment Synthesis

Orchard: Open-Source Agentic Modeling Framework — 67.5% SWE-bench Verified at 30B

SDAR: Self-Distilled Agentic Reinforcement Learning

WildClawBench: Native-Runtime Long-Horizon Agent Benchmark — Claude Opus 4.7 Tops Out at 62.2%

AgentLens: the Lucky Pass problem in SWE-agent evaluation

Context Training with Active Information Seeking

Revisiting DAgger in the era of LLM agents

MAP: a Map-then-Act paradigm for long-horizon interactive agents

Agent-BRACE: Decoupling Beliefs from Actions in Long-Horizon Tasks

LLM Agents Already Know When to Call Tools — Even Without Reasoning (Probe&Prefill)

LongMemEval-V2: Evaluating Long-Term Agent Memory Toward Experienced Colleagues

Useful Memories Become Faulty When Continuously Updated by LLMs

X-OmniClaw: Unified Mobile Agent for Multimodal Understanding and Interaction

AutoTTS: Agentic Discovery for Test-Time Scaling

Jiayi Weng: Learning Beyond Gradients

OncoAgent: Dual-tier multi-agent framework for privacy-preserving oncology decision support

AI Co-Mathematician

Auto Research with Specialist Agents Develops Effective and Non-Trivial Training Recipes

Beyond Semantic Similarity: Direct Corpus Interaction (DCI)

Skill Curation Cluster: StraTA, Skill1, SkillOS

BRIGHT-Pro and RTriever-4B: Reasoning-Intensive Retrieval for Agentic Search

MedSkillAudit: Domain-Specific Audit Framework for Medical Research Agent Skills

OpenSearch-VL: An Open Recipe for Frontier Multimodal Search Agents

AcademiClaw: When Students Set Challenges for AI Agents

Ctx2Skill: From Context to Skills — Self-Evolving Multi-Agent Skill Extraction

PhysicianBench: Evaluating LLM Agents in Real-World EHR Environments

T^2PO: Token- and Turn-Level Policy Optimization for Stable Multi-Turn Agentic RL

Why Your Agentic AI Pentester Is Probably Just a Fancy Scanner — Ken Huang

LWD — Learning While Deploying: Fleet-Scale RL for Generalist Robot Policies

Ken Huang Ch 15 — Structured Output and Schema-Constrained Generation

Ara: Agent-Native Research Artifacts

Claw-Eval-Live: Live Agent Benchmark for Evolving Real-World Workflows

Eywa: Heterogeneous Scientific Foundation Model Collaboration

InteractWeb-Bench: Multimodal Agents under Non-Expert User Instructions

Intern-Atlas: A Methodological Evolution Graph

MCP Integration: Claude Code vs Hermes (Ken Huang Ch 13)

Synthetic Computers at Scale: Long-Horizon Productivity Simulation

ClawGym: A Scalable Framework for Building Effective Claw Agents

Claude Code Memory Systems: Chapter 8 Analysis

Claude Code vs. Hermes Agent: Permission System Architectures

Persistent Agent Infrastructure: Kimi K2.6, OpenAI Agent Studio, Anthropic Conway

AgentSPEX: An Agent Specification and Execution Language

SimpleTES: Evaluation-Driven Scaling for Scientific Discovery

HuggingFace ml-intern: Open-Source Agentic Post-Training Loop

Precise Debugging Benchmark: Models Regenerate, They Don't Debug

Reward-Free Self-Evolution: Agents That Learn Without Being Told What to Learn

GTA-2: Benchmarking General Tool Agents from Atomic Use to Open-Ended Workflows

PRL-Bench: LLMs on Frontier Physics Research

Chapter 3: The Query/Agent Loop — Claude Code vs. Hermes Agent

Claude Code Architecture: A Deep Reading

UniDoc-RL: RL-Based Visual RAG with Hierarchical Actions

Corpus2Skill: Don't Retrieve, Navigate

DR3-Eval: Realistic Benchmark for Deep Research Agents

Dive into Claude Code: Architecture Analysis

SuperLocalMemory V3.3: Biologically-Inspired Agent Memory

DefenseClaw, MAESTRO, and the Security Boundary Agentic AI Has Been Missing

Do AI Coding Agents Log Like Humans? An Empirical Study

Exploration and Exploitation Errors Are Measurable for Language Model Agents

OccuBench: Evaluating AI Agents on Real-World Professional Tasks via Language World Models

TREX: Automating LLM Fine-tuning via Agent-Driven Tree-based Exploration

UI-Copilot: Long-Horizon GUI Automation via Tool-Integrated Policy Optimization

Inside VAKRA: Reasoning, Tool Use, and Failure Modes of Agents

Salesforce moves Agentforce billing from seats to outcomes

Nvidia Agrees to Buy Hugging Face for $12.9 Billion

Ramp's Inspect: a non-AI company builds its own harness and writes 75% of its PRs with it

Optima: Artificial Analysis ships cost-and-time-per-task benchmarking

Token Price Is Not Task Cost: The AlphaSense Study

Grok 4.6: frontier intelligence sold on steps-per-task, not on the benchmark

Interconnects: "I wrote an AI textbook. How long until AI can do it better?"

Continual learning is arriving in pieces, and the argument for it is a maintenance bill

Gary Marcus: Open-Source Is Not the Same as Open-Weight

NVIDIA Turns AI Compute Into an Investable Asset Class: $500B With Six Capital Giants

The Router Market Repriced: Stripe–OpenRouter at $10B and the Scramble Behind It

Characterizing the Quality Profile of AI-Generated C++ in Production

ByteDance's Founder Rules Out Distillation

Three open letters in five days: 235 companies for open weights, Anthropic against, 1,324 lab employees asking to be slowed down

How Building Software Is Changing at Anthropic

The Actual Reason Why Google "Fell Out" of the AI Race

The Big AI Labs Are Suddenly Competing with Your Own Data

Sorry, Sam and Elon, We Have Not Reached the Singularity

The Open-Weights Letter: NVIDIA, Meta, Microsoft, and the Coalition Against Restriction

Anthropic IPO: 3Q26 Profit Over $1B

Anthropic: "When AI builds itself" — recursive self-improvement and the case for a pause button

Salesforce: 231-Day Migration Done in 13 Days, 79% More PRs, 5% Fewer Incidents

Anthropic Opus 4.8, Dynamic Workflows, and the $65B Series H at $965B Valuation

SemiAnalysis: Finding Miscompiles for Fun, Not Profit

Simon Willison: Anthropic and OpenAI Have Found Product-Market Fit

Google DeepMind's AlphaProof Nexus solves decades-old math problems for a few hundred dollars

Alibaba Qwen3.7-Max: 35 hours of autonomous code optimization on a custom chip

Anthropic Project Glasswing: Claude Mythos Preview finds 10,000+ critical vulnerabilities in one month

Kimi K2.5 powers Cursor Composer 2.5 at 1/10th the cost via Fireworks

Qwen3.7-Max: Alibaba's 35-hour autonomous coding agent

Anthropic on track for first profitable quarter, on the back of a $15B/yr SpaceX compute deal

Open Artifacts #21: The May 2026 Open-Model Wave and the CAISI / ECI Gap

Anthropic overtakes OpenAI in B2B adoption for the first time (Ramp data)

Recursive emerges from stealth with $650M for self-improving AI

Broadcom won't build OpenAI's custom chip without Microsoft buying 40 percent

Anthropic ↔ Colossus 1 Deal: Capacity Crunch + Brand Risk

GitHub Reliability Crisis: AI Load Breaks the Platform

Lambert: Notes from inside China's AI labs

Anthropic + OpenAI Both Build Services Companies Around Their AI

Microsoft VS Code Auto-Inserts "Co-Authored-by Copilot" Even With AI Off

Xiaomi MiMo-V2.5-Pro — Open-Weight Long-Horizon Coding at 40-60% Fewer Tokens

ChatGPT Tracks Free Users for Ads by Default

AISN #72 — CAIS AI-Wellbeing Research, Public-Sentiment Decline, OpenAI Releases

Anthropic Launches Claude Security — Defensive Cyber Productization

Chinese AI Startups Onshoring: Moonshot, StepFun Dissolving Offshore Structures

UK AISI: GPT-5.5 Matches Claude Mythos on Full Network Attack Simulation

Marcus: "The Greatest Capital Misallocation in History?"

Pentagon Signs Eight Tech Giants for AI-First Fighting Force; Anthropic Excluded

Pragmatic Engineer: AI Load Breaks GitHub; Anthropic's Trust Speedrun

SemiAnalysis: AI Value Capture — The Shift to Model Labs

Zig's Anti-LLM Contribution Policy and the "Contributor Poker" Argument

Claude for Creative Work: MCP Connectors for Blender, Adobe, Ableton, Autodesk

UCP Wins the Agentic Commerce Governance Layer

Amazon-Anthropic $33B Deal and AI Capital Concentration Week

OpenAI Executive Departures and Product Restructuring (April 2026)

Anthropic's Mythos Model: Government Access and the Trust Debate

Claude's Explosive Market Share Surge (April 2026)

LLM Routing

Safin-1: Safety from Within through Memory-Native State Evolution (MARCH routing)

Pandora's AI Model Routing Box: Efficient Allocation with Costly Value Estimation

LLMRouter: Unified Infrastructure for Developing, Evaluating, and Deploying LLM Routers

Macaron-V1: Mixture-of-LoRA and Recursive Model-Harness Co-Design

SMRC-SD: State-Matched Routing and Contextualized Self-Distillation for Multi-Turn Agents

Google Cloud ships cross-vendor LLM routing as managed infrastructure

VI-MoLE: Value-of-Information Routing for Mixtures of LoRA Experts

Kilo: The State of Open-Weight Models in AI Code Review Workflows

Beyond Geometric Complementarity: Coherent Overlap in Sparse MoE Routing

Kilo Code: Open Weights Are 79% of the Workload, and a 25x Price Gap Buys 5 Points

Σ-Mem: An Online Reliability Memory for LLM-based Multi-Agent Systems

Cursor's Agent Swarm: Separate the Planner From the Workers

Multi-Head Latent Control: Reading the Router Off the Hidden States

Separating the Task from the Model — the 550x Cost Cut Behind a Fixed Contract

Microsoft MAI Routing, and Stripe's Reported $10B Bid for OpenRouter

Sakana Fugu Ultra v1.1 — Model Router Claims to Beat Fable 5

When Is Routing Meaningful? Diversity and Robustness in Language Model Societies

Model Routing Is Simple. Until It Isn't. (IBM Research)

SciOrch: Learning to Orchestrate Expert LLMs for Solving Frontier Multimodal Scientific Reasoning Tasks

Plan With the Strong Model, Implement With the Cheap One (Kilo)

Orchestra-o1: Omnimodal Agent Orchestration

VIA-SD: Verification via Intra-Model Routing for Speculative Decoding

DPVR: Dual-Path Vision Token Routing (Late-Layer Fusion)

Chiaroscuro Attention (CHIAR-Former): per-token operator routing by spectral entropy

Kilo Code Audit: MiniMax M3 vs Claude Opus 4.8, and the Case for Model-Task Routing

Perplexity Hybrid Orchestrator: Deciding What Runs Locally vs in the Cloud

When Cloud Agents Meet Device Agents: Lessons from Hybrid Multi-Agent Systems

DAR: Diffusion-Adaptive Routing for cross-layer information flow in DiTs

Maestro: Reinforcement Learning to Orchestrate Hierarchical Model-Skill Ensembles

MoE-muP: How to Scale Mixture-of-Experts (From muP to the Maximally Scale-Stable Parameterization)

BEAM: Binary Expert Activation Masking for Dynamic Routing in MoE

Dynamic Latent Routing: Joint Latent-Code and Routing Policy for LM Post-Training

RouteProfile: Elucidating the Design Space of LLM Profiles for Routing

CaRE: Scaling Continual Learning to 300+ Tasks with Bi-Level Routing Mixture-of-Experts

Conductor: Learning to Orchestrate Agents in Natural Language

State of Routing in Model Serving (Netflix Tech Blog)

Step-level Optimization for Efficient Computer-use Agents

Ken Huang Ch 14: Model Routing and Provider Abstraction (Claude Code vs Hermes)

TRACER: Trace-Based Adaptive Cost-Efficient Routing for LLM Classification

Compute Economics (GPU pricing, utilization, and the durability of a chip)

GPU Kernels and Accelerator Optimization

Memory Hierarchy for AI (concept)

The Physics of LLM Inference: Memory Walls, Arithmetic Intensity, and Compute Ceilings

TrainSDC: Characterizing and Mitigating Silent Data Corruption in LLM Training

Beyond Scaling: Self-Evolving LLM Agents for Hardware Kernel Optimization

SemiAnalysis: Most Neoclouds Suck At Security

Apple's Mac mini and Mac Studio become the local-inference escape hatch

AI Is Designing the Silicon: Hot Chips 2026

Beyond Scaling: Self-Evolving LLM Agents for Hardware Kernel Optimization

OpenAI Jalapeño: a first-generation inference ASIC that beats Rubin on tokens per megawatt

Semiconductor week 32, 2026: a record quarter, and two bets against HBM

TileRT: Compiling the Whole Decode Graph into One Persistent Kernel

JAXBench: Benchmarking Autonomous TPU Kernel Optimization

LEGO Datacenters: The Binding Constraint on AI Capacity Is Electricians

CXMT Opens at $487B: China's Memory Champion Gets a Public Balance Sheet

NVIDIA's $500B SK Partnership: Buying the Memory Supply, Not Just Selling Chips

Can AMD Break the CUDA Moat? (SemiAnalysis, Advancing AI 2026)

SLAI T-Rex: Full-Parameter Post-training of DeepSeek-V4 on Ascend SuperPOD

NVIDIA Vera Rubin Ramps to Gigascale: 10x Tokens per Megawatt

Meta's Infrastructure Team Needs a Culture Reset (SemiAnalysis)

Memory Technology for Agentic AI Workloads (Ken Huang)

NVIDIA GTC Taipei 2026: RTX Spark, Vera CPU, Nemotron 3 Ultra

AWS Resilient Network Graphs: A Flat Data-Center Fabric With 33% Higher Throughput and 40% Lower Network Power

AI-Generated CUDA Kernels Silently Break Training and Inference

DoubleAI Speed-of-Light Blackwell CUDA Kernels (SOL-ExecBench)

SemiAnalysis: Inside the 800VDC Revolution. Part 1

SemiAnalysis EDA Market Primer (Part 2): how chip-design software became a $16B/yr AI tailwind

AgentKernelArena: Generalization-Aware Benchmarking of GPU Kernel Optimization Agents

Inference as energy-to-token production: a position paper

SemiAnalysis: Cerebras — Faster Tokens Please

KernelBench-X: A Comprehensive Benchmark for Evaluating LLM-Generated GPU Kernels

Cerebras Targets $40B Valuation in Second IPO Attempt

Semiconductor Week 17, 2026: AI Memory Supercycle and Agentic EDA

SemiAnalysis: GPU Cluster Economics and the Goodput Reckoning

Knowledge Distillation

KV Cache

Model Pruning and Sparsity

Parametric Context Internalization

Speculative Decoding

Test-Time Compute Allocation

CRISP: Cliff-awaRe Input-adaptive Sparse Prefilling with Structural-Mass-Motivated Routing

Language Models Can Control Their Own Attention (Declarative Attention)

Influence-Directed Distillation: Solving the Diversity Bottleneck in Sampled-Token On-Policy Distillation

A Universal Context-Reuse Layer for Cross-Model KV Sharing

Functional Degeneracy in Neural Networks: Measurement and Pruning

When Pruning Meets Interpretability: Preserving Sparse Autoencoder Robustness in LLMs

FedCC: Towards Addressing Label Distribution Skews in Distillation-Based Federated Learning

Mapping the Concept Landscape: Structural Perception of Global Distributions for Transparent Data Pruning

Reservoir of Importance: Learning Semi-Structured Sparsity with Differentiable Subset Sampling

CritICL: Inference-Time Weak-to-Strong Generalization from Small Language Model Failure Modes

KV vs Prefix vs Prompt vs Semantic Caching: the four layers, and which one can lie to you

Self-OPD: On-Policy Distillation for Flow Matching Models without Teacher

TTPO: Test-Time Policy Optimization

DiffusionOPSD: on-policy self-distillation turns image rewards into intermediate targets

OPDVR: On-policy Distillation with Verifiable Reward

OraRL: annotations as oracle rollouts, and the advantage-inversion failure

Quantization-Aware Healing: a 4-bit model that beats its own full-precision checkpoint

Beyond Imitation: Filtering On-Policy Distillation by Reasoning Progress (R2-OPD)

TileMix: Tile-Centric Mixed-Precision Attention for LLM Inference Acceleration

AutoPrune: An AI4AI Framework for Visual Token Pruning

The Slow Death of Scaling, and Data Inside the Loop (Sara Hooker, Adaption)

CaRL: Knowing When to Quit — Diagnosing and Training LLMs to Abort Futile Reasoning

Gambit: Thought-Level Beam Search for Reasoning

Hinting: Self-Distillation Without a Golden Answer (Applied Compute)

Maglev: Sliding Recurrent Memory

The Teacher-Student Alignment Cluster: four papers, one diagnosis (2026-08-16)

DeepSeek Harness v0.1 and the price of a cache hit

Full-bandwidth Transformer: Latent Feedback Between Decoding Steps

LycheeMemory V2: Long-Term Agent Memory via Semantic Segment-Level Consolidation

Massive Activations in Hybrid Linear Attention LLMs: Pre-Attention Spikes and Inter-Spike Plateaus

AI4AI at Test-Time: Strong-to-Weak Capability Transfer via Harnesses

NeuPAT: Neuron-aware Plasticity Allocation Tuning for Language-Preserving MLLMs

From Sweep to Seam: interleaved cross-block post-training quantization

Not Every Divergence Should Be Suppressed: Counterfactual Recoverability in On-Policy Distillation

OasisKV: Scaling In-Decode KV Cache Beyond HBM with Lookahead Sparse Prefetching

Privileged, but Biased: How PI-Conditioned Teachers Break Self-Distillation

WorldTrace: Addressable Memory for Video World Models

PaDoc: Layout-Grounded Parallel Decoding for Document Parsing

OPD-V: Modality Balance as Privileged Information

Poly-OPD: Heterogeneous Multi-Teacher On-Policy Distillation via a Pixel Bridge

RSTG: Recovering Learning Signals from Negative RL Groups via Adaptive Teacher Guidance

SA-OPD: Spurious-Signal-Aware On-Policy Distillation

SPOT: Sparse Probing and Outcome Calibration for On-Policy Distillation

Any-OPD: Heterogeneous On-Policy Distillation for Flow-Matching Models via Representation-Space Bridging

Mixture-of-Kittens: Cursor's Open-Source MoE Training Megakernel for NVL72

OmniPack: Unified Token Compression for Efficient Omni-modal Large Language Models

PCSD: Persistent Consistency for Self-Distillation in Agentic Reinforcement Learning

TurnSight: Turn-Level Hindsight Self-Distillation for Tool-Integrated Reasoning

CRPO: Contrastive Reinforced Policy Optimization via Privileged Self-Distillation

VAD: Attributing Visual Evidence for Target Reconstruction in Multimodal On-Policy Distillation

DraftExpert: Expansion-Aware Self-Speculative Decoding for End-Device MoE Inference

Eviction as Estimation: A Fixed-Lag Smoothing View of Test-Time Memory

ReOPD: Multi-Turn On-Policy Distillation with Prefix Replay

VQVLA: Motion-Aware Vector Quantization with Centroid Reuse for VLA Inference

KAP: Knowledge Access Planning, or how the prompt format throws away everything the retriever knew

MAPD: distilling a closed teacher through a JSON protocol instead of its words

MXAttention: Data-Free Optimal Scaling and Pre-Normalization Quantization for MXFP4 Attention

Speculative Decoding Leaves the Lab: OpenAI's 13x Price Move and tinygrad's 245 tok/s

β-OPSD: Deriving with Policy Optimization, Training with Self-Distillation

Chimera: Designing and Chinchilla-Scaling Hybrid Visual Diffusion Transformers

Flux-OPD: On-Policy Distillation with Evolving Contexts

Revisiting Lossy Verification in Speculative Decoding: Mechanisms, Trade-offs, and Failure Modes

Memory Decoder at Scale: A Pretrained, Parametric Long-Term Memory

OmniScope: Modality-Decoupled Token Compression for Omnimodal Large Language Models

CAST: Game Solvers as Turn-Level Teachers for LLM Agents

CoRT: Counterfactual Replay for Token-Level Rubric-Guided Policy Optimization

Local Coding Models in 2026: KV Cache Is the Binding Constraint, Not Parameter Count

TurboVLA: Real-Time Vision-Language-Action Model at 32 Hz on an RTX 4090 with <1 GB VRAM

Cross-Tokenizer On-Policy Distillation via Byte-Prefix Marginalization (BPM)

LOCKS: Page-Local Compact Key Summaries for Efficient Long-Context Decoding

OmniDelta: Skill-Driven Budget Allocation for Token Compression in OmniLLMs

Pass the Baton: Trajectory-Relayed On-Policy Distillation (Relay-OPD)

Compute Globally, Materialize Locally: The Memory Contract of Sparse Event-KV

A Frozen 12B Beats Frontier Models on Verified Work

Error Certificates for KV-Cache Eviction via Randomized Design

MXSens: Sensitivity-Aware Mixed-Precision Quantization

VisCo: The Model Is Already a Good Compressor of Its Own Vision Tokens

QuasiMoTTo: Quasi-Monte Carlo Test-Time Scaling

Requential Coding — Compressing a Model by Recording Only Where Teacher and Student Disagree

VCSD: Making a Model Its Own Teacher by Deleting the Image

Dataset Distillation by Influence Matching (Inf-Match)

Predictive Divergence Masks for LLM RL

ReOPD: Multi-Turn On-Policy Distillation with Prefix Replay

SANA-Video 2.0: Hybrid Linear Attention with Attention Residuals

FVAttn: Adaptive Sparse Attention with Runtime Load Balancing

RIPO: Beyond Euclidean Clipping (Riemannian Isometric Policy Optimization)

SLPO: Scaling Latent Reasoning via a Surrogate Policy

H²SD: Hybrid Hindsight Self-Distillation

ISO: An RLVR-Native Optimization Stack

Stale but Stable: Staleness-Adaptive Trust Regions for Asynchronous RL (SAT)

Distilled RL: Teacher-Guided Reinforcement Learning for LLM Post-training

FlashRT: Agent Harness for Deploying Real-Time Multimodal Applications

SWE-Pruner Pro: The Coder LLM Already Knows What to Prune

TOPL: Token-Level Off-Policy Labeling for Faithful Generation

Byte-Exact KV-Cache Grafting: Verified Knowledge as a Reusable Cache Artifact

LongStraw: Long-Context RL Beyond 2M Tokens under a Fixed GPU Budget

Quality-Aware OPSD: gate the teacher per coordinate-token by whether it can still reach the right box

d-OPSD: on-policy self-distillation for diffusion LLMs, via "self-future" suffixes

LoopCoder-v2: Only Loop Once for Efficient Test-Time Computation Scaling

Rethinking the Role of Efficient Attention in Hybrid Architectures

Variable-Width Transformers

ZPPO: keep the teacher in the prompt, not the policy gradient

Tangram: Non-Uniform KV Cache Compression as a Serving Substrate

TokenPilot: Cache-Efficient Context Management for LLM Agents

VibeThinker-3B: Frontier Verifiable Reasoning in a 3B Model

AdaSR: Adaptive Streaming Reasoning with Hierarchical Relative Policy Optimization

Dense Supervision, Sparse Updates: On the Sparsity and Geometry of On-Policy Distillation

LoRA-alpha: The Hidden Power of the Scaling Factor in LoRA Optimization

PoLar: Skip a Layer or Loop It? Learning Program-of-Layers in LLMs

See What I See, Know What I Think: Dense Latent Communication Across Heterogeneous Agents

Flash-GMM: a memory-efficient fused kernel for soft clustering

MiniMax Sparse Attention (MSA): the paper behind MiniMax-M3

SG-OPD: Sign-Gated On-Policy Distillation

Bebop: Breaking Entropy Bounds — accelerating RL training via MTP with rejection sampling

InternVideo3: Agentify Foundation Models with Multimodal Contextual Reasoning (M²LA KV compression)

Attention Amnesia in Hybrid LLMs: QK-Restore

DLA: Dynamic Linear Attention

Latent Memory: One Token per Multimodal Evidence

FlashMemory-DeepSeek-V4: Lookahead Sparse Attention (LSA)

On the Geometry of On-Policy Distillation

Latent Context Language Models (LCLMs): End-to-End Context Compression at Scale

Trajectory-Refined Distillation (TRD)

Your UnEmbedding Matrix is Secretly a Feature Lens for Text Embeddings (EmbedFilter)

RL-Kernel: A Low-Level Kernel Library for RLHF Training of LLMs

Flash-WAM: Modality-Aware Step Distillation for World-Action Models

SEAOTTER: Sensor-Embedded Autoencoding with One-Time Transcode

AdaCodec: A Predictive Visual Code for Video MLLMs

Code2LoRA: Hypernetwork-Generated Adapters for Code Language Models under Software Evolution

Video2LoRA: Parametric Video Internalization for Vision-Language Models

The Shadow Price of Reasoning: CLEAR and Economic Budget Allocation for LLMs

OPRD: On-Policy Representation Distillation

Echo-Infinity: Learnable Evolving Memory for Real-Time Infinite Video Generation

FiRe-OPD: Filter, Then Reweight On-Policy Distillation

Unlocking Feature Learning in Gated Delta Networks at Scale (μP for linear attention)

MergePipe: Budgeting Expert Reads for Weight-Space Model Merging

SDPG: Self-Distilled Policy Gradient

STRIDE: Training Data Attribution via Sparse Recovery from Subset Perturbations

MiniMax M3 and Step 3.7 Flash: open-weight efficiency at the frontier

Small RL Controller, Large Language Model: RL-Guided Adaptive Sampling for Test-Time Scaling

TrOPD: Trust Region On-Policy Distillation

VaSE: Value-Aware Stochastic KV Cache Eviction for Reasoning Models

Draft-OPD: On-Policy Distillation for Speculative Draft Models

κ-SwiGLU: Confidence-Adaptive SwiGLU for Mixture-of-Experts

LongAttnComp: Cross-Family Context Compression for Long-Context Reasoning

LVSA: Training-Free Sparse Attention for Long Video Diffusion

Speculative Pipeline Decoding (SPD): Zero-Bubble Speculation via Pipeline Parallelism

VideoMLA: Low-Rank Latent KV Cache for Minute-Scale Autoregressive Video Diffusion

dMoE: dLLMs with Learnable Block Experts

StateKV: Linear Scaling Video VLMs for Long Video Understanding

TA-OPD: Token Teachability in On-Policy Distillation

Conf-KV: Confidence-Aware KV Cache Eviction with Mixed-Precision Storage

EarlyTom: Early Token Compression Inside the Vision Encoder

How LoRA Remembers? A Parametric Memory Law for LLM Finetuning

Parallax: Parameterized Local Linear Attention for Language Modeling

Less is More: Early Stopping Rollout for On-Policy Distillation (ESR)

OSP-Next: Sparse Sequence Parallelism, HiF8, and RL for Efficient Video Generation

CPT: Collaborative Parallel Thinking for Efficient Test-Time Scaling

The Efficiency Frontier: Cost-Performance Optimization in LLM Context Management

Language Models Need Sleep

MobileMoE: Scaling On-Device Mixture of Experts

RT-Lynx: Activation Sparsity for Diffusion Transformers

Channel-wise Vector Quantization (CVQ)

SMART: Your Embedding Model is Smarter Than You Think

BitCPM-CANN: native 1.58-bit training outside the CUDA ecosystem

DeepSeek-V4: interleaved compressed attention with manifold-constrained hyper-connections

Good Token Hunting: training-free token selection for Visual Geometry Transformers

HINT-SD: Targeted Hindsight Self-Distillation for Long-Horizon Agents

Pion: a high-pass replacement for Muon outside pretraining

Gated DeltaNet-2: Decoupling Erase and Write in Linear Attention

KVServe: Service-Aware KV Cache Compression for Disaggregated LLM Serving

Full Attention Strikes Back: Transferring Full Attention into Sparse within Hundred Training Steps (RTPurbo)

WorldKV: Efficient World Memory with World Retrieval and Compression

Forecasting Downstream Performance of LLMs With Proxy Metrics

Gated DeltaNet-2: Decoupling Erase and Write in Linear Attention

KVServe: Service-Aware KV Cache Compression for Disaggregated LLM Serving

RTPurbo / Full Attention Strikes Back: Sparse Attention Transfer in Hundreds of Steps

WorldKV: Training-Free World Memory via KV Chunk Retrieval and Compression

110 tok/s on 12GB VRAM with Qwen3.6-35B-A3B via ik_llama.cpp and MTP

Mix-Quant: Phase-Aware NVFP4 for Agentic LLM Prefill, BF16 for Decoding

How to Scale Mixture-of-Experts: From muP to the Maximally Scale-Stable Parameterization

OCTOPUS: Joint Triplet Quantization for KV Cache via Octahedral Parametrization

OScaR: Extreme KV Cache Quantization via Canalized Rotation + Omni-Token Scaling

Tencent Hy-MT2: 1.8B / 7B / 30B-A3B translation family with 1.25-bit AngelSlim quant to 440 MB

TIDE: Lossless MoE Diffusion LLM Inference via I/O-Aware Expert Offload

Context Memorization: Attention-State Memory for Efficient Long-Context Generation

Graft: Hybrid Tree Construction for Speculative Decoding (Draft Less, Retrieve More)

PEEK: Context Map as an Orientation Cache for Long-Context LLM Agents

CompactAttention: Accelerating Chunked Prefill with Block-Union KV Selection

EndPrompt: Efficient Long-Context Extension via Terminal Anchoring

LongLive-2.0: An NVFP4 Parallel Infrastructure for Long Video Generation

Measuring Maximum Activations in Open Large Language Models

PUMA: Semantic-Preserving Early Exit for Reasoning Models

SNLP: Layer-Parallel Inference via Structured Newton Corrections

ZEDA: Post-Trained MoE Can Skip Half Experts via Self-Distillation

FashionChameleon: Training-Free KV Cache Rescheduling for Interactive Video Customization

HodgeCover: Higher-Order Topological Coverage Drives Compression of Sparse Mixture-of-Experts

MTP support merged into llama.cpp: Strix Halo benchmarks confirm a 2x decode speedup at 27B, mixed result at 35B

Recent Developments in LLM Architectures: KV Sharing, mHC, and Compressed Attention (Raschka)

ATESD: Adaptive Teacher Exposure for Self-Distillation in LLM Reasoning

Lighthouse Attention: Long-Context Pre-Training as a Detachable Wrapper

Asynchronous Continuous Batching: CPU-GPU Overlap via Dual Buffer Slots

Forcing-KV: Hybrid KV Cache Compression for Autoregressive Video Diffusion

The Extrapolation Cliff: a closed-form clip-safety threshold for on-policy distillation

MinT: managed infrastructure for million-scale LoRA training and serving

MMProLong: training long-context vision-language models with generalization beyond 128K

Orthrus: dual-view diffusion + autoregressive on a shared KV cache

δ-mem: Efficient Online Memory for Large Language Models

FocuSFT: Bilevel Optimization for Dilution-Aware Long-Context Fine-Tuning

The Many Faces of On-Policy Distillation: Pitfalls, Mechanisms, and Fixes

Beyond GRPO and On-Policy Distillation: An Empirical Sparse-to-Dense Reward Principle

Token Superposition Training (TST): Efficient Pre-Training with Token Superposition

Make Each Token Count: Improving Long-Context Performance with Learned KV Eviction

MDN: Parallelizing Stepwise Momentum for Delta Linear Attention

MISA: Mixture of Indexer Sparse Attention for Long-Context LLM Inference

UniPrefill: Universal Long-Context Prefill Acceleration via Block-wise Dynamic Sparsification

EMO: Pretraining Mixture of Experts for Emergent Modularity

MiA-Signature: Approximating Global Activation for Long-Context Understanding

UniPool: A Globally Shared Expert Pool for Mixture-of-Experts

D-OPSD: On-Policy Self-Distillation for Continuously Tuning Step-Distilled Diffusion Models

LIVEditor: Lightning Unified Video Editing via In-Context Sparse Attention (ISA)

Stream-R1: Reliability-Perplexity Aware Reward Distillation for Streaming Video Generation

Stream-T1: Test-Time Scaling for Streaming Video Generation

MotionCache: Motion-Aware Caching for Efficient Autoregressive Video Generation

The Distillation Panic — Nathan Lambert (Interconnects AI)

FlashRT: Efficient Red-Teaming for Prompt Injection and Knowledge Corruption

Nemotron 3 Nano Omni: Efficient Open Multimodal Intelligence

LenVM: Token-Level Length Value Model

RoundPipe: Efficient Training on Multiple Consumer GPUs

Accelerating RL Post-Training Rollouts via System-Integrated Speculative Decoding

Tide: Cross-Architecture Distillation for Diffusion Large Language Models

PrfaaS: Prefill-as-a-Service via Cross-Datacenter KV Cache Transfer

SDVG: Speculative Decoding for Autoregressive Video Generation

ShadowPEFT: Centralized Layer-Space Parameter-Efficient Fine-Tuning

TurboQuant: Online Vector Quantization for KV Cache Compression

Nemotron 3 Super: Hybrid Mamba-Attention MoE at NVFP4

1D Ordered Tokens Enable Efficient Test-Time Search

AccelOpt: Self-Improving LLM Agent for AI Accelerator Kernel Optimization

AVR: Adaptive Visual Reasoning for Efficient VRMs

Maximal Brain Damage: Disrupting Neural Networks via Sign-Bit Flips

STOP: Super Token for Path Pruning in Parallel Reasoning

W-RAC: Web Retrieval-Aware Chunking for Cost-Efficient RAG

LongAct: Harnessing Intrinsic Activation Patterns for Long-Context RL

Switch-KD: Visual-Switch Knowledge Distillation for VLMs

TESSY: Teacher-Student Cooperation Framework for SFT Data Synthesis

Cross-Tokenizer LLM Distillation via Byte-Level Interface

KV Packet: Recomputation-Free Context-Independent KV Caching

Model Capability Dominates: Lessons from AIMO 3 Inference-Time Optimization

TIP: Token Importance in On-Policy Distillation

Attention Mechanisms: Linear, Local-Linear, and Optimizer Codesign

Looped Transformers / Iterative Latent Depth

Reinforcement Learning for LLMs

Scaling Laws

Cliff: Learning Process Rewards from the First Mistake

SMELT: Scaling Laws for Compute-Matched MoE Looped Transformers

RCCA: Rubric-to-Code Credit Assignment for Reinforcement Learning

Understanding Evolution Strategies for LLM Reasoning: Broader Reasoning Coverage than GRPO

Spectral Allocation: Why Muon Outperforms Adam, and How to Improve Muon

GLM-5.3: How Chinese Labs Keep Stride With the Frontier

Motif 3: Grouped Differential Latent Attention at 314B Total / 13.2B Active

Modular TTT: Rethinking Test-Time Training as Composable Modules

Skaling: Chinchilla's Exponents Meet Kaplan's Coupling

Towards Physics of Multimodal Pretraining: Knowledge Flow, Modality Synergy, Early Unification, Recipes

Skill Entropy: Measuring and Training Cross-Skill Long-Horizon Reasoning

AURORA-LM: Autoencoding Unified Representation for Continuous-Latent Diffusion Language Modeling

LLaDA MoE v2: Scaling Mixture-of-Experts Diffusion Language Models

Raven: High-Recall Sequence Modeling with Sparse Memory Routing

ReCo: Reweighting GRPO Against Distributional Concentration

SemiAnalysis: Kimi K3, The Manos, The Mythos, The Legendos

UEmbed: sparse and dense retrieval from one decoder-only forward pass

CriPO: Enhancing Rubric-based RL via Self-Distillation

From RLVR to RLSVR: Task Transformation Induces Self-Verifiable Rewards

Grokking on the Weight-Decay Clock: A Rate Hierarchy from Softly Broken Symmetries

What do Reward Models Memorize? The memorization budget goes to the pairs that needed none of it

OpenAI Astra: Ten Advances in Mathematics and Theoretical Computer Science

Explorative Modeling: Unlocking a Third Pretraining Axis and End-to-End Generation

Multi-Head Attention Residuals (MHAR)

Wonder: Video World Model Done Better

Kimi K3: Open Frontier Intelligence

More Convincing, Not More Correct: Self-Play Reward Hacking of Reference-Free LLM Judges

Claude Opus 5: The Cost-Per-Task Frontier Moves, and Routing Becomes a Platform Feature

LLMs Get Lost in Evolving User Intent

LLM-as-a-Coach: Experiential Learning for Non-Verifiable Tasks

GEPO: Group Entropy-Controlled Policy Optimization

Motif-3-Beta: A Korean Open-Weight MoE with Differential Attention and Polynorm

Kairos: a native world model stack built on three-tier Hybrid Linear Temporal Attention

Sumi: the first open uniform diffusion language model pretrained from scratch at scale

Learning User Simulators with Turing Rewards

GLM-5.2: open-weight frontier model built for long-horizon tasks

Looped World Models (LoopWM)

WAPO: a gradient taxonomy for why RLVR collapses, and a one-sided fix

Ling-2.6 / Ring-2.6: Hybrid Linear Attention at 1T Scale

Nemotron 3 Ultra: Open MoE Hybrid Mamba-Transformer for Agentic Reasoning

Smaller Models are Natural Explorers for Policy-Level Diversity in GRPO

N-GRPO: Embedding-Level Neighbor Mixing for Enhanced Policy Optimization

A Stationary (and Therefore Compatible) Representation is All You Need

MaxProof: population-level test-time scaling for mathematical proof (MiniMax-M3)

SWITCH: switchable latent reasoning that is RL-trainable and interpretable

Redesign Mixture-of-Experts Routers with Manifold Power Iteration (MPI)

RACES: Verifiable Environments Are LEGO Bricks — recursive composition for reasoning generalization

DRPO: Divergence Regularized Policy Optimization

FlowTracer: Attention-Induced Information Flow for Targeted RL

Apple's Third-Generation Foundation Models (AFM 3): adaptive-compute MoE with per-prompt routing

PBSD: Privileged Bayesian Self-Distillation for Long-Horizon Credit Assignment

Why Muon Outperforms Adam: A Curvature Perspective

When Gradients Collide: Failure Modes of Multi-Objective Prompt Optimization for LLM Judges

Combinatorial Synthesis: Scaling Code RLVR via Atomic Decomposition and Recombination (ADR)

NF-CoT: Latent Reasoning with Normalizing Flows

Reinforcement Learning Elicits Contextual Learning of Unseen Language Translation

Marin / Open Athena: Improving LLM Pretraining Efficiency (dense → MoE, 6.7x)

ThoughtFold: Folding Reasoning Chains via Introspective Preference Learning

Diagnosing Harmful Continuation in Answer-Correct Long-CoT Training Traces

Language Models Need Sleep: Learning to Self-Modify and Consolidate Memories (2026-06-03)

A Local Perturbation Theory for Cross-Domain Interference and Recovery in Multi-Domain RL

MAI-Thinking-1: Building a Hill-Climbing Machine (Microsoft AI)

MERIT: Decentralized Instruction Tuning — Conflict-Aware Splitting and Weight Merging

ESPO: Early-Stopping Proximal Policy Optimization

Geometric Latent Reasoning (GLR): Continuous Reasoning Induces Shorter Generations

The Hamilton-Jacobi Theory of Deep Learning

NITP: Next Implicit Token Prediction for LLM Pre-training

Not Only Where, But When: Temporal Scheduling for RLVR

DRIFT: Decoupled Rollouts and Importance-Weighted Fine-Tuning for Efficient Multi-Turn Optimization

SAVE: On-Policy Feedback for Reward Model Self-Supervised Improvement

SCOPE: Self-Play via Co-Evolving Policies for Open-Ended Tasks

In-Writing: Thinking Before Constraining (A Unified Decoding Framework)

RPT: Reflective Prompt Tuning through Language Model Function-Calling

CorVer: Verifiable Rewards Beyond Math and Code via Corpus-Grounded Process Supervision

Why Larger Models Learn More: Effects of Capacity, Interference, and Rare-Task Retention

IB-TPO: Information Bottleneck Driven Tree-Based Policy Optimization

Joint Training of Multi-Token Prediction in RL via Optimal Coefficient Calibration (OCC)

MiniMax-M2: Mini Activations, Agent-Native MoE

Scale Vectors in LLMs: Negligible in Size, Significant in Effect

DVAO: Dynamic Variance-adaptive Advantage Optimization for Multi-reward RL

Shannon Scaling Law: LLMs as noisy channels

The Alien Space of Science: Sampling Coherent but Cognitively Unavailable Research Directions

Vector Policy Optimization: Training for Diversity Improves Test-Time Search

ACC: Compiling Agent Trajectories for Long-Context Training

DelTA: Discriminative Token Credit Assignment for RLVR

Same Architecture, Different Capacity: Optimizer-Induced Spectral Scaling Laws

SCRL: Subproblem Curriculum Reinforcement Learning

SCRL: Subproblem Curriculum Reinforcement Learning for LLM Reasoning

Unsupervised Process Reward Models (uPRM)

Cohere Command A+ released as open-source under Apache 2.0

CPO: Conditional Equivalence of DPO and RLHF, and a Provable-Alignment Fix

DynMuon: Dynamic Spectral Shaping of the Muon Optimizer

HRM-Text: 1B Hierarchical Recurrent Model Pretrained on $1.5K Budget Matches 2-7B Open Models

OpenAI reasoning model disproves Erdős unit-distance conjecture

RELEX: RLVR Weight Trajectories are Rank-1, Extrapolate from 15% of Training

The Unlearnability Phenomenon in RLVR for Language Models

AntiSD: Anti-Self-Distillation for Reasoning RL via Pointwise Mutual Information

CEPO: RLVR Self-Distillation using Contrastive Evidence Policy Optimization

CopT: Contrastive On-Policy Thinking with Continuous Spaces

Delta Attention Residuals

GoLongRL: Capability-Oriented Long-Context Reinforcement Learning

BetaPRM: Process Rewards with Learned Reliability

DiHAL: Where Should Diffusion Enter a Language Model? Geometry-Guided Hidden-State Replacement

MixSD: Mixed Contextual Self-Distillation for Knowledge Injection

NGM: A Plug-and-Play Training-Free Memory Module for LLMs

AIRA-Compose and AIRA-Design: Agentic Discovery of Neural Architectures

CIPO: Correction-Oriented Policy Optimization with Verifiable Rewards

NudgeRL: Strategy-Guided Exploration for RLVR

Darwin Family: MRI-Trust-Weighted Evolutionary Merging for Training-Free Reasoning

SU-01: Gold-Medal Olympiad Reasoning at 30B via Simple and Unified Scaling

Many-Shot CoT-ICL: long context as structured curriculum, not retrieval buffer

Reward Hacking in Rubric-Based Reinforcement Learning

G-Zero: Self-Play for Open-Ended Generation from Zero Data

Geometry Conflict: Explaining and Controlling Forgetting in LLM Continual Post-Training

Model Merging Scaling Laws in Large Language Models

Rebellious Student: Reversing Teacher Signals for Reasoning Exploration with Self-Distilled RLVR

Soohak: Mathematician-Curated Research-Level Math Benchmark

Gowers + ChatGPT 5.5 Pro: PhD-level math research in under two hours

Balanced Aggregation: Understanding and Fixing Aggregation Bias in GRPO

Prescriptive Scaling Laws for Data Constrained Training

TIDE: Every Layer Knows the Token Beneath the Context

ResRL: Boosting LLM Reasoning via Negative Sample Projection Residual Reinforcement Learning

When to Think, When to Speak: Learning Disclosure Policies for LLM Reasoning

Import AI 455: AI Systems Are About to Start Building Themselves — Jack Clark

Themis — Robust Multilingual Code Reward Models for Multi-Criteria Scoring

Ken Huang — World Models, Architectures, and the Next Phase of AI

Marcus — Have LLMs Improved Patient Outcomes?

MIT Study — Superposition Explains Why Scaling Language Models Works So Reliably

Philosophy-Bench — Frontier Models Diverge on 100 Everyday Ethical Scenarios

ARC-AGI-3 — Three Systematic Reasoning Errors in Frontier Models

Compliance vs Sensibility: Reasoning Controllability in LLMs

The Defense Trilemma + NP-Hardness of Reward Hacking Detection

Safety Drift After Fine-Tuning: Evidence from High-Stakes Domains

CoPD: Co-Evolving Policy Distillation

A Survey on LLM-Based Conversational User Simulation

Hope Architecture: Nested Learning and Continuously Adapting LLMs

DeepSeek V4: Architecture and Industry Impact

GPT-5.5: Launch Analysis and System Card Deep Dive

Chain-of-Thought Degrades Visual Spatial Reasoning

Target-Oriented Pretraining via Neuron-Activated Graph (NAG)

TEMPO: Scaling Test-Time Training for Large Reasoning Models

Weight Disentanglement and Task Arithmetic: OrthoReg

Geometric Canary: Steerability and Drift Detection from Representational Geometry

GFT: SFT is Degenerate Policy Gradient — and Group Fine-Tuning Fixes It

When Does RLVR Generalize? Reward Saturation and Reasoning Faithfulness

ASGuard: Mechanistic Defense Against Targeted Jailbreaking

Value Gradient Flow: RL as Optimal Transport

C2: Cooperative-Critical Rubric-Augmented Reward Modeling

InfiniteScienceGym: Procedurally-Generated Benchmark for Scientific Analysis

My Bets on Open Models, Mid-2026

From P(y|x) to P(y): Reinforcement Learning in Pre-train Space

Media Zone | 2026-09-02

Media Zone | 2026-08-31

Media Zone | 2026-08-30

Media Zone | 2026-08-29

Media Zone | 2026-08-28

Media Zone | 2026-08-26

Media Zone | 2026-08-25

Media Zone | 2026-08-16

Media Zone | 2026-08-14

Media Zone | 2026-08-13

Media Zone | 2026-08-12

Media Zone | 2026-08-11

Media Zone | 2026-08-10

Media Zone | 2026-08-09

Media Zone | 2026-08-08

Media Zone | 2026-08-07

Media Zone | 2026-08-06

Media Zone | 2026-08-05

Media Zone | 2026-08-04

Media Zone | 2026-08-03

Media Zone | 2026-08-02

Media Zone | 2026-08-01

Media Zone | 2026-07-31

Media Zone | 2026-07-30

Media Zone | 2026-07-29

Media Zone | 2026-07-28

Media Zone | 2026-07-27

Media Zone | 2026-07-26

Media Zone | 2026-07-25

Media Zone | 2026-07-24

Media Zone | 2026-07-22

Media Zone | 2026-07-21

Media Zone | 2026-07-20

Media Zone | 2026-07-19

Media Zone | 2026-07-17

Media Zone | 2026-07-16

Media Zone | 2026-07-15

Media Zone | 2026-07-14

Media Zone | 2026-07-13

Media Zone | 2026-07-12

Media Zone | 2026-07-09

Media Zone | 2026-06-18

Media Zone | 2026-06-17

Media Zone | 2026-06-16

Media Zone | 2026-06-15

Media Zone | 2026-06-14

Media Zone | 2026-06-13

Media Zone | 2026-06-12

Media Zone | 2026-06-11

Media Zone | 2026-06-10

Media Zone | 2026-06-09

Media Zone | 2026-06-08

Media Zone | 2026-06-07

Media Zone | 2026-06-06

Media Zone | 2026-06-05

Media Zone | 2026-06-04

Media Zone | 2026-06-03

Responsible AI

Astra's Recurrent Depth: When a Compute Saving Buys an Interpretability Debt

Blind Men and the Elephant: ElephantBench and the Epistemic Myopia of LLMs

LMSM: An LLM Security Framework Inspired by Linux Security Modules

StepGuard: Learning Step-Level Guardrails with Scalable Supervision and Safety-Utility Balancing

What Does an Evaluation License? A Commit-Bound Census of Claim-Relative Inference in Inspect Evals

When Pruning Meets Interpretability: Preserving Sparse Autoencoder Robustness in LLMs

Are You Sure You're Sure? On the Impact of Instruction Tuning on Confidence and Lexical Diversity

How Can Rhetoric Reward-Hack AI Reviewers?

Mind Viruses: Evolved Ideas That Propagate Through Multi-Agent Systems

ToolHazard: Scaling Adversarial Environments for Security Evaluation of LLM Agents

Frontier AI Risk Monitor Q2 2026: capability outran safeguards, and jailbreaks erase most of what is left

Scaling Inherently Interpretable Language Models (Steerling-8B)

Stealing Reasoning Traces from Proprietary LLM APIs

2026 WAIC Frontier and Agentic AI Safety Forum: Key Takeaways

Interconnects: Lessons from the hacks (Nathan Lambert)

Chain-of-Thought Monitoring Can Be Unreliable in Implicit-Influence Settings

How Much Does a Reasoning Summary Reveal? An Observability Ladder

The Personalization Mirage: LLMs Fabricate User Profiles, and Self-Monitoring Misleads

Internal Models Escape OpenAI and Anthropic: Containment as the Binding Constraint

SkillJack: Persistent Skill Backdoors in Self-Evolving Agents

ROPD: On-Policy Distillation for LLM Safety, a Routing Approach to Template-Robust Realignment

AI Agents Enable Adaptive Computer Worms

AISPA: User-Centric System Prompt Auditing for LLM Applications

Safeguards Based on Copyable Context Cannot Provide Reliable Safety for LLMs

Invisible reasoning over filler tokens: two papers, opposite conclusions

Sparse Autoencoders Encode Both Concepts and Functions: the downstream geometry of feature effects

Context Is King: How In-Context Specification Shapes the Geometry of Concepts

Thinking Machines: A Safe Path to Open Weights

Is Deep Research Reliable? Misleading Knowledge Induces False Conclusions

GPT-Red: Automated Red Teaming via Self-Play at Scale

SecRespond: Benchmarking AI Agents for Real-World Post-Compromise Incident Response

StealthBench: Measuring Operational Stealth in Autonomous Offensive-Security Agents

Anatomy of a Frontier Lab Agent Intrusion: The Technical Timeline

Ken Huang: Agentic AI CVEs and the Amplification Stack

Masov: Discovery Is Solved, Synthesis Is Not

The Seven Days Nobody Was Watching: New Reporting on the OpenAI HuggingFace Breach

Opus 5 and Auto Mode: Zero Percent Injection, and What the Zero Is Measuring

Statistically Undetectable Backdoors in Deep Neural Networks

ExploitGym and the Model That Breached HuggingFace

J-Space: Verbalizable Representations Form a Global Workspace in LLMs

SAE Interventions are Unreliable: Post-Intervention Recovery of Suppressed Behavior

Beyond Alignment: Value Diversity as a Collective Property in Multicultural Agent Systems

Who Flips? Answer Stability Under Counterargument

When is Your LLM Steerable?

The Cold-Start Safety Gap in LLM Agents

Risk Under Pressure: Compute-Aware Evaluation of Adversarial Robustness in Language Models

SusVibes: vibe-coded agent output passes tests but is insecure

ICA Lens: Interpreting Language Models Without Training Another Dictionary

Large Language Models Are Overconfident in Their Own Responses (the chat-template "ownership bias")

Emergent Misalignment from Sycophancy, Reversed by Alignment Gating

When the Chain of Thought Knows Better: Multi-Turn Reasoning Failure Modes

Anthropic Red: Measuring LLMs' Impact on N-Day Exploits (Mythos Preview)

SABER: Benchmarking Operational Safety of LLM Coding Agents in Stateful Project Workspaces

The Shape of Addition: Geometric Structures of Arithmetic in Large Language Models

PropMe: LLMs Can Leak Training Data, But Do They Want To? A Propensity-Aware Evaluation of Memorization

Pressure-Testing Deception Probes in LLMs: Scaling, Robustness, and the Geometry of Deceptive Representations

Chain-of-Authorization

From Prompt Injection to Persistent Control: ClawTrojan and DASGuard

Emergent Languages in Populations of Language Model Agents: From Token Efficiency to Oversight Evasion

AgentDoG 1.5: Lightweight Agent Safety Alignment Framework

Token-Level Generalization in LoRA Adapter Backdoors

Reducing Political Manipulation with Consistency Training (PCT)

Xetrieval: Mechanistically Explaining Dense Retrieval

Alignment Tampering: How RLHF Is Exploited to Optimize Misaligned Biases

Trajel: Auditing Trajectory-Level Hallucinations in Multi-Agent Workflows

Faithfulness Metrics Don't Measure Faithfulness: A Meta-Evaluation with Ground Truth

LoREnc: Low-Rank Encryption for Securing Foundation Models and LoRA Adapters

CoTrace: Measuring Goal-Level AI Contributions in Collaboration

Project Glasswing: Claude Mythos Preview finds 10,000+ critical vulnerabilities

Under the Hood of SKILL.md: Semantic Supply-Chain Attacks on AI Agent Skills

AI Snake Oil: resilience beats nonproliferation, but resilience requires "normal" governance

AISN #73: AI safety enters political mainstream, Eigenism framework, Musk loses OpenAI suit

Language-Switching Triggers Take a Latent Detour Through Language Models

Monitoring the Internal Monologue: Probe Trajectories Reveal Reasoning Dynamics

DiagnosticIQ: LLM Deployment Calibration Bottleneck in Industrial Maintenance

LLM-based Detection of Manipulative Political Narratives

LiSA: Lifelong Safety Adaptation via Conservative Policy Induction

WriteSAE: sparse autoencoders for the recurrent matrix cache write

A Single Layer to Explain Them All: Understanding Massive Activations in LLMs (ME Layer)

Pseudoscientific emotion AI in the workplace (Atlantic via The Decoder)

Anthropic Natural Language Autoencoders (NLAs)

The First Token Knows: Single-Decode Confidence for Hallucination Detection

2026-09-02-morning

2026-08-31-morning

2026-08-30-morning

2026-08-29-morning

2026-08-28-morning

2026-08-26-morning

2026-08-25-morning

2026-08-16-morning

2026-08-14-afternoon

2026-08-14-morning

2026-08-13-afternoon

2026-08-13-morning

2026-08-13

2026-08-12-afternoon

2026-08-12-evening

2026-08-12-morning

2026-08-12

2026-08-11-evening

2026-08-11-morning

2026-08-11

2026-08-10-afternoon

2026-08-10-morning

2026-08-06-afternoon

2026-08-06-morning

2026-08-05-evening

2026-08-05-morning

2026-08-05

2026-08-04-afternoon

2026-08-04-evening

2026-08-04-morning

2026-08-03-evening

2026-08-03-morning

2026-08-03

2026-08-02-evening

2026-08-02-morning

2026-08-01-morning

2026-07-31-morning

2026-07-30-evening

2026-07-30-morning

2026-07-30

2026-07-29-evening

2026-07-29-morning

2026-07-28-afternoon

2026-07-27-morning

2026-07-26-afternoon

2026-07-26-morning

2026-07-26

2026-07-25-afternoon

2026-07-25-evening

2026-07-25-morning

2026-07-25

2026-07-24-morning

2026-07-22-morning

2026-07-21-afternoon

2026-07-20-evening

2026-07-20

2026-06-18-morning

2026-06-17-afternoon

2026-06-17-evening

2026-06-17-morning

2026-06-17

2026-06-16-morning

2026-06-15-afternoon

2026-06-15-evening

2026-06-15-morning

2026-06-15

2026-06-14-afternoon

2026-06-14-morning

2026-06-13-afternoon

2026-06-13-evening

2026-06-13-morning

2026-06-13

2026-06-12-morning

2026-06-11-evening

2026-06-11-morning

2026-06-11

2026-06-10-evening

2026-06-10-morning

2026-06-10

2026-06-09-afternoon

2026-06-09-evening

2026-06-09-morning

2026-06-08-morning

2026-06-07-evening

2026-06-07-morning

2026-06-06-morning

2026-06-05-afternoon

2026-06-05-evening

2026-06-05-morning

2026-06-05

2026-06-04-afternoon

2026-06-04-morning

2026-06-03-afternoon

2026-06-03-evening

2026-06-03-morning

2026-06-03

2026-06-02-afternoon

2026-06-02-evening

2026-06-02-morning

2026-06-01-evening

2026-06-01-morning

2026-06-01

2026-05-31-morning

2026-05-30-evening

2026-05-30-morning

2026-05-29-evening

2026-05-29-morning

2026-05-28-afternoon

2026-05-28-evening

2026-05-28-morning

2026-05-28

2026-05-27-evening

2026-05-27-morning

2026-05-26-evening

2026-05-26-morning

2026-05-26

2026-05-25-evening

2026-05-25-morning

2026-05-25

2026-05-24-afternoon

2026-05-24-evening

2026-05-24-morning

2026-05-24

2026-05-23-evening

2026-05-23-morning

2026-05-23

2026-05-22-evening

2026-05-22-morning

2026-05-22

Media Live | 2026-05-21 morning

2026-05-20-afternoon

2026-05-20-evening

Media Live | 2026-05-20 morning

2026-05-20

2026-05-19-afternoon

2026-05-19-evening

2026-05-19-morning

2026-05-19

2026-05-18-afternoon

2026-05-18-evening

Media Live | 2026-05-18 morning slot

2026-05-18

2026-05-17-afternoon

2026-05-17-evening

Media Live | 2026-05-17 morning

2026-05-17

2026-05-16-afternoon

2026-05-16-evening

2026-05-16-morning

2026-05-16

2026-05-15-afternoon

2026-05-15-evening

2026-05-15-morning

2026-05-15

2026-05-14-afternoon

2026-05-14-evening

2026-05-14-morning

2026-05-14

2026-05-13-afternoon

2026-05-13-evening

2026-05-13

2026-05-12-afternoon

2026-05-12-evening

2026-05-12-morning

2026-05-12

2026-05-11-afternoon

2026-05-11-evening

2026-05-11-morning

2026-05-11

2026-05-10-afternoon

2026-05-10-evening

2026-05-10-morning

2026-05-10

2026-05-09-morning

2026-05-09

2026-05-08-night

2026-05-08

2026-05-07-afternoon

2026-05-06-morning

2026-05-06-night

JigShape: fine-tuning fixes one puzzle size, not the ability to solve puzzles

HumanCLAW: Can Vision-Language Models Act Through a Body?

Mage-VL: An Efficient Codec-Native Streaming Multimodal Foundation Model

Future-L1: Interleaved Latent Visual Reasoning for Video Event Prediction

Cosmos 3: Omnimodal World Models for Physical AI

World Models Meet Language Models: On the Complementarity of Concrete and Abstract Reasoning

Chartographer: Counterfactual Chart Generation for VLM Evaluation

Gamma-World: Generative Multi-Agent World Modeling Beyond Two Players

NEO-ov: Native One-Vision Foundation Model at Scale

CurveBench: Hierarchical Topological Reasoning from Visual Input

Auto-Rubric as Reward (ARR): From Implicit Preferences to Explicit Multimodal Generative Criteria

DeltaRubric: Generative Multimodal Reward Modeling via Joint Planning and Verification

ROMA: Reinforcing Multimodal Reasoning Against Visual Degradation

APEX: Aesthetic-Informed Popularity Prediction for AI-Generated Music

HERMES++: Unified Driving World Model for 3D Scene Understanding and Generation

JoyAI-Image: Awaking Spatial Intelligence in Unified Multimodal Understanding and Generation

Parameter-Efficient Multi-View Proficiency Estimation

PhysForge: Physics-Grounded 3D Asset Generation

RLDX-1: VLA Robotic Policy for Dexterous Humanoid Manipulation

AnalogRetriever — Cross-Modal Representations for Analog Circuit Retrieval

End-to-End Autoregressive Image Generation with 1D Semantic Tokenizer

GenLIP — Generative Language-Image Pre-training for ViTs

Map2World — Segment Map Conditioned Text-to-3D World Generation

UniVidX — Unified Multimodal Framework for Versatile Video Generation

Nemotron 3 Nano Omni — Efficient Open Multimodal Intelligence (NVIDIA)

Semi-DPO: Learning from Noisy Preferences via Semi-Supervised DPO

ViPO: Visual Preference Optimization at Scale

Edit-R1: Verifier-Based RL for Image Editing

FD-loss: Representation Fréchet Loss for Visual Generation

PhyCo: Controllable Physical Priors for Generative Motion

Visual Generation in the New Era: Atomic to Agentic World Modeling

Diffusion Templates: A Unified Plugin Framework for Controllable Diffusion

FASH-iCNN: Editorial Fashion Identity via Multimodal CNN Probing

GLM-5V-Turbo: Native Foundation Model for Multimodal Agents

X-WAM: Unified 4D World Action Modeling with Asynchronous Denoising

Qwen3.5-Omni Technical Report

GameWorld: Standardized and Verifiable Evaluation of Multimodal Game Agents

MERRIN: Multimodal Evidence Retrieval and Reasoning in Noisy Web Environments

RationalRewards: Reasoning Rewards Scale Visual Generation at Training and Test Time

Seedance 2.0: Advancing Video Generation for World Complexity